What happened
RAND researchers evaluated 12 U.S. Department of War sites offered for AI data center leases against 17 Department of Energy sites and built a quantitative energy-cost model across 31 candidate locations, finding that 'across both departments, nearly every candidate site lacks sufficient existing generation capacity to power a gigawatt-scale data center without significant new infrastructure.' At gigawatt scale, energy infrastructure costs rival data center construction costs, and a stakeholder 'Day After' exercise found that community opposition and infrastructure disputes could escalate into 'overlapping legal, political, and security crises.' The report proposes a site-screening framework ranking candidates on energy costs, grid interconnection, and infrastructure readiness, and urges agencies to treat energy as a 'coequal workstream' alongside civil construction in lease solicitations.
Why it matters
With the U.S. government opening federal land for AI infrastructure, this gives executives and policy leads an evidence-based rubric for where (and whether) gigawatt-scale AI data centers are actually viable — energy, not acreage, is the binding constraint.
Action needed
Map the RAND siting framework to any federal land or hyperscale site-selection pipeline your organization is assessing, and stress-test energy-independent cost assumptions before committing capital.